THE MONIST all Cretans were liars. It can be put more simply in the form: if a man makes the statement I am lying, is he lying or not? If he is, that is what he said he was doing, so he is speaking the truth and not lying. If, on the other hand, he is not lying, then plainly he is speaking the truth in saying that he is ly-ing, and therefore he is lying, since he says truly that that is what he is doing. It is an ancient puzzle, and nobody treated that sort of thing as anything but a joke until it was found that it had to do with such important and practical problems as whether there is a greatest cardinal or ordinal number. Then at last these contradictions were treated seriously. The man who says I am lying is really asserting There is a proposition which I am asserting and which is false. That is presumably what you mean by lying. In order to get out the contradiction you have to take that whole assertion of his as one of the propositions to which his assertion applies; i.e. when he says There is a proposition which I am asserting and which is false, the word proposition has to be inter-preted as to include among propositions his statement to the effect that he is asserting a false proposition. Therefore you have to suppose that you have a certain totality, viz., that of propositions, but that that totality contains members which can only be defined in terms of itself. Because when you say There is a proposition which I am asserting and which is false, that is a statement whose meaning can only be got by reference to the totality of propositions. You are not saying which among all the propositions there are in the world it is that you are asserting and that is false. Therefore it presup-poses that the totality of proposition is spread out before you and that some one, though you do not say which, is being asserted falsely. It is quite clear that you get a vicious circle if you first suppose that this totality of propositions is spread out before you, so that you can without picking any definite

356

THE PHILOSOPHY OF LOGICAL ATOMISM one say Some one out of this totality is being asserted false ly, and that yet, when you have gone on to say Some one out of this totality is being asserted falsely, that assertion is itself one of the totality you were to pick out from. That is exactly the situation you have in the paradox of the liar. You are supposed to be given first of all a set of propositions, and you assert that some one of these is being asserted falsely, then that assertion itself turns out to be one of the set, so that it is obviously fallacious to suppose the set already there in its entirety. If you are going to say anything about all propositions, you will have to define propositions, first of all, in some such way as to exclude those that refer to all the propositions of the sort already defined. It follows that the word proposition, in the sense in which we ordinarily try to use it, is a meaningless one, and that we have got to divide propositions up into sets and can make statements about all propositions in a given set, but those propositions will not themselves be members of the set. For instance, I may say All atomic propositions are either true or false, but that it-self will not be an atomic proposition. If you try to say All propositions are either true or false, without qualification, you are uttering nonsense, because if it were not nonsense it would have to be itself a proposition and one of those in-cluded in its own scope, and therefore the law of excluded middle as enunciated just now is a meaningless noise. You have to cut propositions up into different types, and you can start with atomic propositions or, if you like, you can start with those propositions that do not refer to sets of proposi-tions at all. Then you will take next those that refer to sets of propositions of that sort that you had first. Those that refer to sets of propositions of the first type, you may call the sec-ond type, and so on. If you apply that to the person who says I am lying,

357

THE MONIST tions excepting for accidental and irrelevant linguistic forms, with, however, a certain proviso which must now be ex-plained.

Take, e.g. two propositional functions such as x is a man, x is a featherless biped. Those two are formally equivalent, i.e. when one is true so is the other, and vice versa. Some of the things that you can say about a proposi-tional function will not necessarily remain true if you substi-tute another formally equivalent propositional function in its place. For instance, the propositional function x is a man is one which has to do with the concept of humanity. That will not be true of x is a featherless biped. Or if you say, so-and-so asserts that such-and-such is a man the propositional function x is a man comes in there, but x is a featherless biped does not. There are a certain number of things which you can say about a propositional function which would be not true if you substitute another formally equivalent propo-sitional function. On the other hand, any statement about a propositional function which will remain true or remain false, as the case may be, when you substitute for it another formally equivalent propositional function, may be regarded as being about the class which is associated with the propo-sitional function. I want you to take the words may be re-garded strictly. I am using them instead of is, because is would be untrue. Extensional statements about functions are those that remain true when you substitute any other formally equivalent function, and these are the ones that may be regarded as being about the class. If you have any statement about a function which is not extensional, you can always derive from it a somewhat similar statement which is ex- tensional, viz., there is a function formally equivalent to the one in question about which the statement in question is true. This statement, which is manufactured out of the one you started with, will be extensional. It will always be equal-

360

THE PHILOSOPHY OF LOGICAL ATOMISM ly true or equally false of any two formally equivalent func-tions, and this derived extensional statement may be regard-ed as being the corresponding statement about the associated class. So, when I say that The class of men has so-and-so many members, that is to say There are so-and- so many men in the world, that will be derived from the statement that x is human is satisfied by so-and-so many values of x, and in order to get it into the extensional form, one will put it in this way, that There is a function formally equivalent to x is human, which is true for so-and-so many values of x. That I should define as what I mean by saying The class of men has so-and-so many members. In that way you find that all the formal properties that you desire of classes, all their formal uses in mathematics, can be obtained without supposing for a moment that there are such things as classes, without supposing, that is to say, that a proposition in which symbolically a class occurs, does in fact contain a constitu-ent corresponding to that symbol, and when rightly analysed that symbol will disappear, in the same sort of way as de-scriptions disappear when the propositions are rightly ana-lysed in which they occur.

There are certain difficulties in the more usual view of classes, in addition to those we have already mentioned, that are solved by our theory. One of these concerns the null-class, i.e. the class consisting of no members, which is difficult to deal with on a purely extensional basis. An-other is concerned with unit classes. With the ordinary view of classes you would say that a class that has only one member was the same as that one member. That will land you in terrible difficulties, because in that case that one member is a member of that class, namely, itself. Take, e.g. the class of Lecture audiences in Gordon Square. That is obviously a class of classes, and probably it is a

361